How many games will a college football team win in the upcoming season? How likely are they to win their conference championship?
This notebook displays 2022 predictions and historical data for Arizona.
Using each team’s starting rating, I simulate the result and margin of victory for each game of the 2022 season 1000 times.
The following table shows the percentage of times in which Arizona won each of the games on their upcoming schedule. The margin of victory is the average margin of victory across all simulations.
Season | Week | Date | Team | Opponent | Prob.Win | Pred.Margin |
2022 | 1 | Sep 03 | Arizona | @ SDSU | 0.163 | -16.0 |
2022 | 2 | Sep 11 | Arizona | vs MSST | 0.330 | -8.5 |
2022 | 3 | Sep 18 | Arizona | vs NDSU | 0.438 | -3.5 |
2022 | 4 | Sep 24 | Arizona | @ CAL | 0.223 | -13.0 |
2022 | 5 | Oct 01 | Arizona | vs COLO | 0.517 | 1.0 |
2022 | 6 | Oct 08 | Arizona | vs ORE | 0.254 | -13.0 |
2022 | 7 | Oct 15 | Arizona | @ WASH | 0.287 | -10.0 |
2022 | 9 | Oct 29 | Arizona | vs USC | 0.436 | -3.0 |
2022 | 10 | Nov 05 | Arizona | @ UTAH | 0.071 | -26.0 |
2022 | 11 | Nov 12 | Arizona | @ UCLA | 0.154 | -21.0 |
2022 | 12 | Nov 19 | Arizona | vs WSU | 0.366 | -8.0 |
2022 | 13 | Nov 25 | Arizona | vs ASU | 0.395 | -6.0 |
It’s important to note that within each simulation the results of each game are not independent. The the (simulated) results of each week update each team’s rating for the following week, which will then influence how likely they are to win in following weeks. This means that a shocking upset early on in a simulation can have ramifications for other teams on the remaining schedule.
The table above indicates the aggregated result of every simulated game for the selected team. The visualization below shows the path of the team’s rating in every simulation. This illustrates the most common paths a team’s season is likely to take by showing how the result of one matchup tends to affect the trajectory of the rest of the season.
The simulations for this upcoming season make use of an (adjusted) Elo rating, which is a rating assigned to each team based on its historical wins and losses.
The following visualizations shows Arizona’s Elo rating from 1900 to present.
In addition to the team’s historical Elo rating, I also compute and use each team’s offensive/defensive efficiency metrics from recent seasons in simulating upcoming games. These ratings come from an expected points model I trained on on play by play data in order to identify the value of individual plays within a game.
I use this model to determine the expected points added from every individual play and then aggregate these results to the game and season level, adjusting for opponent quality. These results are only available starting from the season of 2007, as this is when reliable play by play data is available.
For more details on what goes into creating these ratings, see my description of the expected points model and my methodology for adjusting for opponent quality.
SEASON | TEAM | OFFENSE | DEFENSE | OVERALL |
2007 | Arizona | 0.037 | 0.046 | 0.083 |
2008 | Arizona | 0.098 | 0.023 | 0.121 |
2009 | Arizona | 0.033 | 0.063 | 0.095 |
2010 | Arizona | 0.067 | 0.047 | 0.113 |
2011 | Arizona | 0.107 | -0.087 | 0.020 |
2012 | Arizona | 0.177 | -0.001 | 0.176 |
2013 | Arizona | 0.106 | 0.075 | 0.181 |
2014 | Arizona | 0.086 | 0.026 | 0.112 |
2015 | Arizona | 0.091 | -0.093 | -0.002 |
2016 | Arizona | 0.072 | -0.092 | -0.020 |
2017 | Arizona | 0.156 | -0.056 | 0.100 |
2018 | Arizona | 0.001 | -0.044 | -0.043 |
2019 | Arizona | -0.026 | -0.038 | -0.063 |
2020 | Arizona | -0.127 | -0.075 | -0.202 |
2021 | Arizona | -0.058 | -0.091 | -0.149 |
These efficiency ratings indicate a team’s expected points per play when its offense or defense is on the field, adjusted for opponent quality. A team’s overall rating is a combination of its offense and defense ratings, and indicates the net points per play a team would expect when playing an average opponent.
For offenses, this indicates the average points the team scores against opponents per play. For defenses, this indicates the average points the team prevents opposing offenses from scoring per play. In both cases, I have set the scale of the variable to mean that positive is good while negative is bad.
The visualization below shows how a team has changed over time as well as their ranking for each metric at the end of each season.
In addition to examining each team’s overall efficiency on offense and defense, I can break down each team’s efficiency based on the play type. How has Arizona performed when running/passing on offense vs defending the run/pass on defense?
SEASON | TEAM | OFFENSE_PASS | OFFENSE_RUN | DEFENSE_PASS | DEFENSE_RUN |
2007 | Arizona | 0.055 | 0.037 | 0.082 | 0.096 |
2008 | Arizona | 0.096 | 0.071 | 0.082 | -0.050 |
2009 | Arizona | 0.031 | 0.027 | 0.186 | -0.021 |
2010 | Arizona | 0.175 | 0.032 | 0.132 | 0.004 |
2011 | Arizona | 0.235 | -0.009 | -0.096 | -0.093 |
2012 | Arizona | 0.254 | 0.169 | -0.068 | -0.014 |
2013 | Arizona | 0.055 | 0.203 | 0.107 | 0.108 |
2014 | Arizona | 0.038 | 0.073 | -0.001 | 0.057 |
2015 | Arizona | 0.104 | 0.086 | -0.192 | -0.035 |
2016 | Arizona | 0.008 | 0.126 | -0.085 | -0.058 |
2017 | Arizona | 0.178 | 0.149 | -0.063 | -0.110 |
2018 | Arizona | 0.045 | 0.001 | -0.082 | -0.001 |
2019 | Arizona | -0.037 | 0.014 | -0.001 | -0.069 |
2020 | Arizona | -0.294 | 0.004 | -0.045 | -0.082 |
2021 | Arizona | -0.170 | -0.035 | -0.089 | -0.001 |
The following visualization shows how efficient Arizona’s offense has been running/passing the ball each season since 2007.
The following visualization shows how efficient Arizona’s defense has been in stopping the opponent’s run/pass in each season since 2007.